Project

General

Profile

Production 5 (Neutrino 2020 analysis campaign) Reconstruction book keeping

A record of the new/improved features in the production 5 processing campaign, production philosophy, and bookkeeping of software release history and relevant datasets.

Release notes quasi-up to date as of 27th Nov 2019

  • Current production release R19-11-18-prod5reco.f

Production philosophy

Production will start from scratch (ie new Event Generation and Simulation) - see the reasons for this at the end of the Prod4Tag Wiki.

Reconstruction Improvements for Prod5

Many algorithms were improved for the Production 5 Campaign. Technotes have been written documenting the improvements and training parameters used.
  • Slicer
  • Event CVN
  • Prong CVN
  • 2D Prong CVN
  • Regression CVN
  • ReMId
  • NuMu Cosmic Rejection
  • Nue Cosmic Rejection

R19-11-18-prod5reco release series

Outstanding issues / planned features

none

R19-11-18-prod5reco.a

Notes and bug fixes

The first release of the Production 5 campaign. It was cut from develop, as the remove-bpf-vector branch was merged into develop after the MiniProduction campaign was completed.

R19-11-18-prod5reco.b

Notes and bug fixes

Fixed a bug in the Bad Channel List

R19-11-18-prod5reco.c

Notes and bug fixes

Large swathe of changes primarily focused on CVN training and evaluation.
  • Weights directory removed from the CosRej package.
  • Fix a broken Var for recording fraction of cluster hits used in map.
  • Update Event CVN model files
  • Add code to allow three versions of Event CVN to be ran
  • Update art code to load in multiple versions of Event CVN with the correct module labels.

R19-11-18-prod5reco.d

Notes and bug fixes

  • Correct the version of cvn (Event CVN) UPS Product to be setup.
  • Fix CAFMaker fhicl bugs for CosmicCVN, TrueEnergy, 2D Prong CVN, NC CosRej.
  • Removing unnecessary Prong CVN branches which were still hangovers from Prod4.

Relevant datasets

  • ND Data
  • ND Monte Carlo
  • FD MC

R19-11-18-prod5reco.e

Notes and bug fixes

  • Bump the version of calibration to v14.05, so that ND P1 Data could be ran over.
  • Fix a rare error seen in ND MRDiF processing, to ignore 2D showers.
  • Removing unnecessary Prong CVN branches which were still hangovers from Prod4.
  • Quiet some logs

R19-11-18-prod5reco.f

Notes and bug fixes

  • Bump the version of calibration to v14.06, to fix a typo in v14.05.

Relevant datasets

  • ND Data - draining from R19-11-18-prod5reco.d - for Period1 and DiF failures.

R19-11-18-prod5reco.g

Notes and bug fixes

  • A variety of small changes, mostly added fhicl configurations, to support MRBrem and MRDiF processing in the FD.
  • Fix CAFMaker diblock masks for cosrej files
  • Generally, get FD reco with cosmic filtered files working

R19-11-18-prod5reco.h

Notes and bug fixes

  • Some fixes to MRE
  • Bumped calibcsvs to v14.08 (v14.07 was skipped due to unexpected disk quota issue)

R19-11-18-prod5reco.i

Notes and bug fixes

  • Fixes to BadChannels so they will work correctly with cosmic filtered files, specifically initialization when hits aren't found
  • New calibcsvs

R19-11-18-prod5reco.j

Notes and bug fixes

  • Merge r40521 from trunk. Add a flag to disable filling the HitNu information used in semantic segmentation. This is a significant speed-up in the ND MC reconstruction.
  • Reduce number of CVN outputs to prevent garbage in diff_cafs

R19-11-18-prod5reco.k

Notes and bug fixes

  • New calibrator configurations for light level systematics
  • New calibrator configurations for calibration systematics
  • Minor fhicl fixes to CAFMaker for FD CRY MC and NuMI data processing

R19-11-18-prod5reco.l

Notes and bug fixes

  • Fixed typo in Calibrator fhicl
  • Add calibration shape calibrator configurations

R19-11-18-prod5reco.m

Notes and bug fixes

  • Memory reductions and speed improvements
  • Data overlay-related changes
  • Support for time variation systematic in calibrator and ImprovedTransport
  • Updates to diff_cafs
  • Limit CPUs accessed by TensorFlow
  • Remove nocosmics network
  • Expose PEOnly option for PixelMapProducer to reduce memory

Relevant datasets

  • Most FD Numi data

R19-11-18-prod5reco.n

Notes and bug fixes

  • Major MRE updates, which cause trouble for numu energy downsteram from here.

R19-11-18-prod5reco.o

Notes and bug fixes

  • FD rock exposure accounting
  • Allow CAFMaker to continue with no slices
  • Add a filter to check for FLShits

Relevant datasets

  • FD Rock MC

R19-11-18-prod5reco.p

Notes and bug fixes

  • Bad Channels to v23.0

Relevant datasets

  • Period 9 FD Numi Data

R19-11-18-prod5reco.q

Notes and bug fixes

  • Bad channels to v24.0

Relevant datasets

  • Period 10 FD Numi Data

R19-11-18-prod5reco.r

Notes and bug fixes

  • Final MRE timing fixes
  • Add positrons in RHC MRE

Relevant datasets

  • Replacement FD Numi data cafs
  • Draining of all FD Numi data

R19-11-18-prod5reco.s

Notes and bug fixes

  • Add extractHDF5Metadata

R19-11-18-prod5reco.t

Notes and bug fixes

  • Get extractHDF5Metadata into bin

R19-11-18-prod5reco.v

Notes and bug fixes

  • Experimental fix to HDF5 metadata json string
  • Adding file_name to HDF5Maker_module.cc

R19-11-18-prod5reco.w

Notes and bug fixes

  • Fixing FileType bug in HDF5Maker

R19-11-18-prod5reco.x

Notes and bug fixes

  • Fix for missing DataTier in h5 metadata

R19-11-18-prod5reco.y

Notes and bug fixes

  • Adding a truth continament filter for use in the Single Particle sample.
  • Adding the CAFMaker and HDF5Maker modules into the Prod5 SP fhicl.

R19-11-18-prod5reco.z

Notes and bug fixes

  • PEOnly flag to allow making h5 from ben singles pid

R19-11-18-prod5reco.za

Notes and bug fixes

  • Fill wsid for WS CNN output

R19-11-18-prod5reco.zb

Notes and bug fixes

  • Option for running single output wrong sign network

R19-11-18-prod5reco.zc

Notes and bug fixes

  • Adding wscnn to externals setup script